To make evidence-based policy on well-being, policy-makers need information on what causes well- being and on how well-being affects their other objectives (e.g. physical health). Both needs could be satisfied if we had adequate causal models of the development of well-being, physical health etc over the life-course. Policy-makers could then use these models to help evaluate the long-term effects of different types of intervention at different ages - selecting those that gave the biggest improvement to well-being (or other objectives) per net dollar spent. (The model itself helps us to estimate the financial savings resulting from the expenditures.) To take forward this agenda we have four specific aims. Aim 1. Using standard birth cohort surveys, we shall estimate a path-model of well-being over the life course. There will be eight state variables: well-being; conduct/criminality; physical health; cognitive achievement; and (as adults) employment; welfare receipt; income; and family functioning. Each variable at each age will be regressed on previous values of all variables, background variables and on shocks. The British Cohort Study (BCS) and UK National Child Development Study (NCDS) will enable us to estimate these models up to middle age. But these data have the problem that one variable at age t may be related to the same variable at an earlier age (or another variable), without the relationship being causal. The two variables may be determined by a third omitted variable, including most obviously genetic endowment and other unmeasured characteristics of family background. Using identical twins reared together enables us to control for these. Aim 2. Using twin surveys from the UK, Norway and Sweden, we shall estimate similar models to before - once using the data on who is twinned with whom, and once not using it. This will give us a good idea about how far the models in (1) above overestimate the true causal links. (Twin data are not sufficient to replace the earlier data altogether.) Aim 3. Using surveys of ageing, we shall complement (1) with models covering the later years of life. The focus will be especially on how well-being affects physical health and vice versa. Attention will be paid to the moderating effects of social class and social support and to the impact of specific life transitions such as retirement and bereavement. The surveys used will be mainly the English Longitudinal Survey of Ageing, the Whitehall II study and the U.S. Health and Retirement Survey. Aim 4. Using data on well-being interventions, we shall show how these can be used together with the models already estimated to estimate the long-term benefits and savings from these interventions. Policy- makers will be increasingly interested in the long-term benefits from different possible well-being interventions (and in the savings which could result in reduced costs of health care, social care, criminality and so on). But many intervention studies only tell us the short-run effects. Long-term models can be used to simulate longer-term effects. (Where long-term follow-ups are available we can usefully compare the rate at which their effects fade with the rate of fading obtained from the models we estimate.) The measures of well-being used in these surveys vary but most include for adults life-satisfaction and malaise, and for children emotional well-being and conduct. One feature of the research will be the interaction between emotional well-being/malaise and conduct/criminality. The programme will be a fully-integrated collaboration between economists, psychologists, epidemiologists and psychiatrists.